OncologyLive® is the flagship publication of OncLive®. Each issue delivers expert insights and clinical oncology news and updates, which impact the day-to-day clinical practice of oncology professionals who work tirelessly to help improve patient outcomes.

Technology has rapidly expanded to touch nearly every aspect of health care. This expansion includes machine learning technology, which is quickly emerging as a dynamic aid for clinicians in every aspect of cancer care, from diagnosis to treatment decisions.

Machine learning is a branch of artificial intelligence made up of algorithms that can make predictions based on exposure to data. The capabilities of machine learning models are continually being fine-tuned and improved as computing power and the amount of available data continue to quickly advance.

“One way to make predictions, not just in science but anywhere, is through correlations,” John F. McDonald, PhD, said in an interview with OncologyLive®. “[Machine learning technology] can look for correlations between a molecular biomarker and whether a patient has cancer or not. The same thing applies to [determining] what the optimal therapy is for an individual patient. You can examine the genomic profiles of the tumors of patients who have responded well to chemotherapy vs the genomic profiles of individuals who have not, looking for correlations. Machine learning is looking for correlations embedded within very huge data sets.”

McDonald is a professor in the Parker H. Petit Institute for Bioengineering and Bioscience at Georgia Tech University in Atlanta.

McDonald went on to explain that machine learning models work by analyzing features, which are the individual components of interest that are predetermined by the computer scientist who created the model. Examples of features used in oncologic machine learning models include metabolic profiles, genes, and bases of DNA, McDonald explained.

There are several different computational approaches that can be used in a machine learning- based model to aid in making predictions and treatment decisions in cancer care, including linear models, decision trees, and ensemble methods.

Linear models determine risk using a weighted linear combination of patient features. Decision trees divide the feature space into subgroups with similar outcome predictions and the ensemble approach generates predictions using several individual trees.

Applying Machine Learning Techniques to Community Practice

Machine learning will never completely replace the insights and experience of a human clinician, but it is increasingly becoming a valuable asset for clinicians in several areas of oncology. The approach has been tested in many aspects of cancer care and has produced impressive results that warrant further investigation and adoption by clinicians.

McDonald was a part of a team of investigators who evaluated a machine learning-based model known as the ‘ensemble-based machine-learning algorithm (ELAFT).’ The system combined 4 algorithm types to make single predictions.

In the study, investigators sought to predict which of 7 chemotherapeutic agents would be the most effective therapy for a given cancer type. The predictive model was applied to 15 cancer types using data from 499 independent cell lines. The chemotherapies included were carboplatin, cisplatin, docetaxel, doxorubicin, gefitinib, gemcitabine, and paclitaxel.

The model was evaluated based on accuracy, precision, and recall. Accuracy was defined as the fraction of cancer cell lines the algorithm correctly predicted a response for compared with the actual response to the agent observed in other studies. Precision was determined by the fraction of cell lines the model predicted as sensitive to a given agent divided by all cell lines it predicted as sensitive. Recall was calculated by determining the proportion of cell lines the model accurately predicted as sensitive to response to an agent among all the cell lines that were sensitive.

The model achieved an accuracy of 92.38% across cancer types. Recall was also very high at 98.21% and the precision figure was 85.52%. Investigators validated the results by comparing them to a clinical data set of the 7 chemotherapeutic agents, either alone or in combination, in 23 patients with ovarian cancer. The ovarian cancer data were provided by the Ovarian Cancer Institute.

“This kind of machine learning approach could be very useful in generating a decision tool to help physicians,” McDonald said. “Even though it is not going to necessarily be 100% accurate; it is only as good as the data set you put in it. At this point in time, we do not fully understand the cause-and-effect relationships in cancer. I see machine learning as an interim solution that could give us some predictive accuracy, even if we do not understand the underlying cause-and-effect relationships.”

Machine learning technology has also been used to effectively predict the metastatic status of patients with cancer. Green et al recently presented data showing that the ensemble machine learning–based predictive model XGBoost significantly outperformed an approach that relied solely on the International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM) metastasis diagnosis codes in terms of predicting the metastatic status of patients with breast cancer or lung cancer.

Investigators used data from 28,043 patients, pooled from certified tumor register abstraction and electronic health records of 2 large health systems. The patients were split into training (n = 22,434) and testing (n = 5609) groups. XGBoost was then ‘trained’ by analyzing approximately 750 features of the patient data in the training group, which were collected at the time of diagnosis or after.

For patients with breast cancer who were found to have a negative metastatic status, the predictive model achieved 97% precision and 98% recall compared with 91% and 94%, respectively with the ICD-10-CM diagnosis codes. Among patients with lung cancer with a negative metastatic status, the precision and recall were 83% and 90% vs 66% and 93% with the machine learning model and ICD-10-CM codes, respectively (TABLE

Patients with breast cancer who were found to have a positive metastatic status experienced a significant benefit in precision and recall with the machine learning model compared with the ICD-10-CM approach; 90% vs 68% and 86% vs 58%, respectively. Benefits with XGBoost were also seen among patients with lung cancer with a positive metastatic status: Precision and recall were 89% and 81%, respectively, compared with 87% and 52% achieved with the ICD-10-CM approach.

Study authors concluded that the approach could be applied to other cancer types and is scalable. In the future, the model could also be applied to smaller, time-sensitive pieces of data that could be used by clinicians to help determine if and when a patient’s metastatic status has changed.

Additionally, mortality prediction has become an area of interest for investigators looking to apply machine learning methods. Ye et al implemented a machine learning model into community practice setting capable of predicting the 90-day mortality risk of patients with metastatic cancer. Clinicians then used these predictions to attempt to facilitate advance care planning earlier than when it would have occurred without the aid of the tool, ultimately increasing prognostic awareness and hospice enrollment.

The study evaluated the model in 889 patients from 5 participating practices. Seven practices comprising 774 patients included in the study did not implement the machine learning tool and served as the control arm. The median age in both arms of the study was 74 years and the median metastatic lengths were approximately the same at 249 and 250 days, respectively.

Patients in the 5 participating practices were scored every 2 weeks to gauge mortality risk. The machine learning system then used this data, as well as additional data from electronic medical records and historical claims, to evaluate the need for advance care planning in each patient.

The machine learning tool was tested over an 11-month span and was able to generate 10,910 unique predictions from each patient, including 1663 high-risk predictions.

The weighted mean of advance care planning use was 34.4% (range, 19.4%-55.8%) in the participating practices compared with 14.0% (range, 7.4%-31.0%) in the nonparticipating practices, which was a 2.5-fold change in favor of the intervention (

Investigators noted that implementation of the tool is an ongoing effort and that more practices have already planned to adopt it.

They also wrote that the intervention’s impact on hospice enrollment, emergency department visits, and hospital admission will be evaluated in future studies.

Challenges Remain Despite Some Robust Findings

McDonald noted that machine learning in cancer care is still very much in its infancy and significant barriers remain to the widespread adoption of the approach. One of the major challenges is the availability of reliable data needed to make accurate predictions. Data must be accurately entered into electronic health records and these records must have the capability to be easily shared on a large scale to ensure maximum accuracy and efficacy of machine learning approaches. Although small data sets can be useful, machine learning is more effective in finding correlations in large sets.

Additionally, clinicians cannot assume that predictions obtained with machine learning are always 100% accurate. At times, the models can identify patterns and subgroups within data where there is no clear predictive outcome. Predictions should always be reviewed and double-checked for clinical validity.

The amount of information required in electronic health records to make machine learning effective could also become burdensome for clinicians. Delegation of administrative responsibilities will be key to prevent burnout in this area.

Machine learning may also present a substantial learning curve for some clinicians, although the technology is increasingly becoming more integrated and intuitive. However, the benefits of machine learning approaches will likely outweigh the shortcomings and it will continue to emerge as a powerful complementary tool in cancer care.

“[Machine learning] is going to get better as we get more and more data,” McDonald concluded. “As it gets better, clinicians will start adopting it and utilizing it. As it becomes clear that [machine learning] is very accurate, it could become a decision tool [that is used] right off the bat. But that is going to take a while. That is not going to happen right away.”

Nagy M, Radakovich N, Nazha A. Machine learning in oncology: what should clinicians know? 

. 2020;4:799810. doi:10.1200/CCI.20.00049

Bertsimas D, Wiberg H. Machine learning in oncology: methods, applications, and challenges. J

. 2020;4:885-894. doi:10.1200/CCI.20.00072

Lanka J, Housley SN, Benigno BB, McDonald JF. ELAFT: an ensemble-based machine-learning algorithm that predicts anti-cancer drug responses with high accuracy. 

. 2021;4(1):1-11. doi:10.31829-2637-6148.jor2021-4(1)-111

Green F, Huang HT, Lerman M, et al. Using machine learning on real-world data to predict metastatic status. 

. 2022;40(suppl 16):1550. doi:10.1200/JCO.2022.40.16_suppl.1550

Ye P, Butler BM, Vo DM, et al. The initial outcome of deploying a mortality prediction tool at community oncology practices. 

. 2022;40(suppl 16):1521. doi:10.1200/JCO.2022.40.16_suppl.1521

Articles

The expansion of technology includes machine learning, which is quickly emerging as a dynamic aid for clinicians in every aspect of cancer care, from diagnosis to treatment decisions.

Machine Learning Develops Into  a Powerful Tool in Cancer Care

As the therapeutic paradigm for oncology moves toward targeted therapies, drug development has also undergone a shift toward advancing precision medicine in specific patient populations. In parallel, the exponential growth of computing technology and the widespread availability of electronic health data have permitted real-world data (RWD) to play an ever-increasing role in advancing cancer research and bringing therapies to patients who need them, especially in rare cancers with high unmet need. 

 article “Real-World Evidence Should Be Part of Clinical Trials in Oncology,” by Sarah Alwardt, PhD, was a call to action for the industry to rethink how we gather data on the day-to-day usefulness of drugs for regulatory decisions.1 In this article, we will discuss how a growing acceptance of the value of real-world evidence (RWE) has led to its expanded use in oncology and explore opportunities for RWE to support drug development and regulatory decisions moving forward.

According to FDA definitions, RWD relate to patient health status, or the delivery of health care routinely collected from a variety of sources, including electronic health records (EHR), billing and claims databases, and disease registries. RWE is the clinical evidence regarding the usage and potential benefits, or risks, of a medical product derived from analysis of RWD.

RWE complements clinical trials by capturing data on the day-to-day use of drugs and can be used to support both regulatory and policy decisions. Although randomized control trials (RCT) remain the gold standard for evaluating drug safety and efficacy, they are expensive, time-consuming, and often conducted among relatively homogenous patient populations, limiting their generalizability to broader patient populations. Further, the homogeneity of RCT populations may exacerbate health disparities as diverse patients are relatively underrepresented in RCTs.

With the ability to track more patients over a longer period, real-world studies complement clinical trials by providing actionable insight into how treatments perform in clinical practice. The FDA is predicting that greater use of RWE will result in safety and efficacy information becoming available sooner in the postmarketing setting and help to further inform regulatory decisions.

The 21st Century Cures Act, enacted in 2016, requires the FDA to develop a framework for evaluating RWE in the context of drug regulation including to support the approval of new agents and new indications for previously approved drugs, and to support post-approval study requirements.

In September 2021, the FDA published the fourth of a series of industry guidance documents in response to the Cures Act mandate, titled “Real-World Data: Assessing Electronic Health Records (EHRs) and Medical Claims Data to Support Regulatory Decision-Making for Drug and Biological Products.”

 This document is important for drug innovators because it reflects the FDA’s stance on how to narrow the gap between the methodological rigor of conducting RCTs vs using RWE data for the purposes of generating evidence to support regulatory submissions.

In RCTs, a control group demonstrates outcomes for patients with similar characteristics but who do not receive the experimental treatment. External controls, also known as synthetic controls, use an external data source to provide a patient cohort that can be compared against an interventional therapy. 

External control groups serve the same function as traditional control arms but can be used in settings where randomization is infeasible or impractical.

 For example, in rare cancers, it can be difficult to obtain sufficient patient numbers for a traditional control arm. According to the FDA, external control arms could help augment new clinical data by allowing biopharma sponsors to reduce the number of patients assigned to the control arm in a randomized trial, or to conduct smaller randomized trials.

Using external controls to evaluate the ongoing performance of drugs in clinical practice provides the ability to identify longer-term benefits as well as adverse events that might not occur during the trial period. This better informs benefit-risk assessments that can be used to support regulatory decisions, demonstrate value, aid in reimbursement, and develop treatment practice strategies and guidelines.

Following are 3 examples of how using RWD have improved our understanding and treatment of patients with cancer.

First FDA Approval of a First-line Therapy for Merkel Cell Carcinoma

APPROACH: Clinical trial eligibility criteria was applied to a real-world setting and identified patients in EHR to serve as a historical control.

SUMMARY: Metastatic Merkel cell carcinoma (mMCC) is an uncommon yet aggressive skin cancer. Despite incidence of mMCC dramatically increasing in past 20 years, development of an evidence-based standard therapeutic regimen was lagging. However, with a small patient population and short survival times, performing a traditional RCT was not feasible. In a study performed through The US Oncology Network, using Ontada’s iKnowMed oncology EHR, a substantial patient population diagnosed and treated for mMCC was identified. Response and duration were determined using RECIST 1.1 criteria for evaluating image study reports. These data were used as a comparator for the sponsored phase 2 study evaluating avelumab (Bavencio).

RESULTS: By using EHR data, investigators were able to conduct a real-world retrospective analysis of historical treatment therapy responses. These results were used in the interpretation of the outcomes from a single-arm clinical trial that suggested that immune checkpoint inhibitors improve treatment outcomes for patients with mMCC. These results helped to establish the f irst FDA approval of a first-line therapy for this patient population.

Supportive Data for Daratumumab Split First Dose for Patients With Multiple Myeloma

APPROACH: The aim of this retrospective observational cohort study was to provide real-world insight into the MM patient population receiving daratumumab (Darzalex) in the community oncology setting and compare the infusion time and safety associated with split first dose.

SUMMARY: Daratumumab, a human immunoglobulin G kappa monoclonal antibody, is approved for the treatment of patients with multiple myeloma (MM). However, 46% of patients experienced an infusion-related reaction during the initial infusion. To mitigate this risk, the first daratumumab infusion was administered more slowly than subsequent infusions, with a median infusion duration of greater than 7 hours.12 To make the first infusion duration more manageable, The US Oncology Network developed a split-dose schedule to administer the first infusion over 2 consecutive days. 

RESULTS: Real-world evidence demonstrated comparable safety and efficacy of the administration of daratumumab as a single 8-hour infusion of the drug vs two 4-hour infusions over 2 days. There were no statistically significant differences in reactions between groups. Although the total administration time was longer among patients receiving a split first dose, the 2-day schedule with a shorter infusion schedule offers a viable treatment option for outpatient treatment.

Confirmed Results of Breast Cancer Therapy in Real-World Setting

APPROACH: The purpose of this retrospective, observational study was to investigate the clinical use of pertuzumab (Perjeta) in a population of patients with HER2-positive metastatic breast cancer (mBC) treated in a real-world community oncology practice setting.

SUMMARY: Pertuzumab (Perjeta), a humanized monoclonal antibody that binds to the extracellular domain of HER2, is approved for use in the firstline setting for patients with HER2-positive mBC based on findings from the pivotal CLEOPATRA (NCT00567190) trial. Results demonstrated that the addition of pertuzumab to standard therapy improved patient outcomes in this setting. 

RESULTS: The safety and progression-free survival (PFS) associated with the addition of pertuzumab to standard therapy in the first-line setting in the real-world setting were consistent with those observed in the CLEOPATRA trial. The ability to confirm results of a randomized trial in real clinical practice increases confidence in the benefits of such intervention.

Expanding Opportunities for RWE in Oncology

The FDA has accepted RWE to monitor and evaluate the safety of drugs in postmarket studies and there is growing acceptance of the value of using RWE to support pre- and postapproval regulatory submissions. Ultimately, the goal of incorporating RWE into clinical research is to improve our understanding of how drugs perform in broader patient groups and to streamline and accelerate the clinical trial and drug development processes to bring breakthrough treatments to patients who need them.

As robust, highly accurate RWD sources continue to be accepted as valid and reliable data sets, the more we hope they will be used in RWE and external control studies to support a variety of regulatory decisions, including the following:

RWE has been proven effective at helping generate hypotheses for testing in RCTs, identifying drug development tools (including biomarker identification), assessing trial design, and assembling external cohorts for uncommon disease subsets.

 Incorporation of RWE has the potential to improve the recruitment of patients as well as speed the development process for new oncology drugs and make them available to patients earlier.

In single-arm trials, the use of external control data can potentially support the development of comparative benchmarks, especially for conditions with high unmet need. If preliminary clinical evidence for an anticancer agent in a single-arm trial suggests a substantial treatment effect, evaluating outcomes in similar groups of patients using RWD may provide a reliable assessment of the safety and effectiveness for comparison. Oncology agents with a breakthrough therapy designation program are especially appropriate candidates for the use of RWD to support regulatory decisions.

The ability to use external controls to replicate and confirm the results of RCTs in clinical practice expands the clinical data associated with a treatment option and increases confidence in the benefits of that intervention.

Oncology therapies approved for a specific indication are frequently used for the treatment of related malignancies. RWE may rigorously document safety and effectiveness with the level of quality and detail needed to support label expansion. Using high-quality RWE to secure broader indications can accelerate access to patients, as well as reduce the costs associated with RCTs.

RWE is playing an increasingly important role in oncology where treatment decisions are affected by histology of disease and for uncommon conditions for which there are no approved treatments or treatments with only conditional approvals. When used as supportive data, RWE can influence drug development and inform regulatory approvals of breakthrough medicines designed to identify unmet needs and improve patient care. By offering value throughout the drug development cycle, RWE can help biopharma companies shorten development time, reduce the cost of clinical trials, inform clinical practice, and clarify the value of new therapies.

Alwardt S. Real-world evidence should be a part of clinical trials in oncology. OncLive®. March 28, 2022. Accessed June 20, 2022. bit.ly/3y7Pj5C

Framework for FDA’s real world evidence program. FDA. December 2018. Accessed June 20, 2022. bit.ly/3OuwPld

Statement from FDA Commissioner Scott Gottlieb, MD, on FDA’s new strategic framework to advance use of real-world evidence to support development of drugs and biologics. News release. FDA. December 06, 2018. Accessed June 20, 2022. bit.ly/3bkA8gs

21st Century Cures Act. FDA. Updated January 31, 2020. Accessed June 20, 2022. bit.ly/3QAioOt

Real-world data: assessing electronic health records and medical claims data to support regulatory decision-making for drug and biological products. FDA. Updated December 10, 2021. Accessed June 20, 2022. bit.ly/3xNhtS2

Thorlund K, Dron L, Park JJH, Mills EJ. Synthetic and external controls in clinical trials – a primer for researchers. Clin Epidemiol. 2020;12:457-467. doi:10.2147/CLEP.S242097

Hatswell AJ, Baio G, Berlin JA, Irs A, Freemantle N. Regulatory approval of pharmaceuticals without a randomised controlled study: analysis of EMA and FDA approvals 1999-2014. BMJ Open. 2016;6(6):e011666. doi:10.1136/bmjopen-2016-011666

 R, Caubel P, Azoulay L, Dreyer NA. Trial designs using real-world data: the changing landscape of the regulatory process. Pharmacoepidemiol Drug Saf. 2020;29(10):1201-1212. doi:10.1002/pds.4932

Beckers F, Capra W, Cassidy A, et al; Friends of Cancer Research Working Group. Characterizing the use of external controls for augmenting randomized control arms and confirming benefit. Presented at: Friends of Cancer Research Annual Meeting 2019. November 12, 2019; Washington, DC. Accessed June 20, 2022. bit.ly/3b6H8gY

Cowey CL, Mahnke L, Espirito J, Helwig C, Oksen D, Bharmal M. Real-world treatment outcomes in patients with metastatic Merkel cell carcinoma treated with chemotherapy in the USA. Future Oncol. 2017;13(19):1699-1710. doi:10.2217/fon-2017-0187

Rifkin R, Singer D, Aguilar KM, Baidoo B, Maiese EM. Daratumumab split first versus single dosing schedule among patients with multiple myeloma treated in a US community oncology setting: a retrospective observational study. Clin Ther. 2019;41(5):866-881.e7. doi:10.1016/j.clinthera.2019.03.013

Darzalex. Prescribing Information. Janssen Biotech; 2022. Accessed June 20, 2022. bit.ly/3N9XOSb

Robert NJ, Goertz HP, Chopra P, et al. HER2-positive metastatic breast cancer patients receiving pertuzumab in a community oncology practice setting: treatment patterns and outcomes. Drug Real World Outcomes. 2017;4(1):1-7. doi:10.1007/s40801-016-0102-5

Real-world evidence is finding an expanded role in oncology, prompting experts to rethink how data are gathered on the day-to-day usefulness of drugs for regulatory decisions.

Expanding Opportunities for  Real-World Evidence in Oncology 

 gene fusions across multiple cancer types have led investigators to confirm the marker in real-world data analyses.

 fusion has become quite famous over the past few years, and we know this is a chromosomal rearrangement that drives oncogenesis in pediatric and adult cancers,” Carsten Bokemeyer, MD, said in a recent presentation of findings from a matching-adjusted indirect comparison (MAIC) of real-world data. “And there are specific inhibitors, such as larotrectinib, that selectively and potently block the activity from this fusion protein and have a high activity in patients with cancer.”

Larotrectinib (Vitrakvi) is a kinase inhibitor indicated for the treatment of adult and pediatric patients with solid tumors with an 

 gene fusion without a known acquired resistance mutation, are metastatic or where surgical resection is likely to result in severe morbidity, or for whom no satisfactory alternative treatments exist or have disease progression following treatment.

In an analysis presented at the 2022 American Society of Clinical Oncology Annual Meeting, larotrectinib demonstrated longer overall survival (OS) compared with standard-of-care (SOC) therapy in patients with TRK fusionpositive cancer.

MAIC is a common methodology that balances patient population characteristics allowing for cross-study comparisons in the absence of individual patient data, Bokemeyer said. The need for MAIC to assess the efficacy of larotrectinib in this patient population is because neurotrophic-TRK (

) fusions are rare in patients, meaning that single-arm trials conducted to evaluate larotrectinib are limited in comparing its effectiveness with SOC regimens.

In their MAIC OS analysis, researchers found a 78% lower risk of death for patients treated with larotrectinib (n = 85) compared with SOC (n = 28), with a hazard ratio of 0.09 before weighting, but after adjusting it was 0.22 (95% CI, 0.09-0.52; 

 = .001). Moreover, after weighting, the median OS of patients on larotrectinib was 39.7 months compared with 10.2 months for patients on SOC treatment.

“We would need further data to confirm this, but this is one of the several analyses now performed in recent years that helps to add some information on how much the benefit of using NTRK fusion inhibition in this specific patient population really is,” said Bokemeyer, who is director of the Division of Oncology, Hematology, and Bone Marrow Transplantation at the University Hospital Hamburg-Eppendorf in Hamburg, Germany.

Investigators developed the 2 study cohorts leveraging insights from a previous analysis from the European Society of Medical Oncology that identified 

 fusion–positive patients who did not receive 

 inhibitors and used this as the SOC cohort by using data from the Flatiron Health/Foundation Medicine clinico-genomic database.

 Then, researchers employed MAIC to compare larotrectinib with SOC by balancing various population characteristics among several studies. Patient-level data was collected from clinical trials with the TRK inhibitor: a phase 1 trial (NCT02122913), the SCOUT trial (NCT02637687), and the NAVIGATE trial (NCT02576431). They then aggregated SOC data by matching individual data on common baseline characteristics.

A log-rank test of equality, to assess if the 2 groups were similar before the initiation of larotrectinib, found no difference between them (

 = .31). After this, an estimation of the treatment effect of larotrectinib vs SOC on OS was conducted.

To assess the OS results with more certainty, researchers also conducted 2 sensitivity analyses as excessive matching variables in MAIC can reduce the effective sample size (ESS) and the larotrectinib follow-up was longer than the SOC follow-up.

To address the high uncertainty in the analysis of base cases, they conducted several analyses relaxing missing data assumption and removing variables that had minimal overlap between the studies. In comparison to the base case, with an ESS of 13.1 and an HR of 0.22, when looking at the missing values in prior lines of therapy and patients’ ECOG performance score, the ESS was 23.6 with an HR of 0.10 (95% CI, 0.04-0.22). Then removing patients with cancer of unknown primary (CUP) tumor and breast cancer, the ESS was 26.4 (HR, 0.10; 95% CI, 0.05-0.21), removing patients with colorectal cancer (CRC) showed the ESS at 37.9 (HR, 0.12; 95% CI, 0.05-0.26), and removing patients with salivary cancer, the ESS ended at 38.6 (HR, 0.12; 95% CI, 0.05-0.25).

A restricted mean survival time analysis was also conducted up to 30.5 months, which was the minimum of the largest observed event time within the SOC arm. The researchers found an estimated 10.8-month mean survival advantage with larotrectinib when compared with SOC (95% CI, 7-14.6; 

 < .01). On average, this meant that potential patients treated with larotrectinib at 30.5 months of follow-up would survive an additional 10.8 months, compared with SOC. However, the researchers wrote in their poster that the analysis was conservative and that the mean benefit would likely continue to increase with a longer follow-up of patients.

Eighty-two percent of patients were positive for 

 fusion in both arms after adjustments, and 39.3% were either 65 years old or older on the trial. In both arms there was an even split between patients with an ECOG performance score of 0 to 1 and 2 to 4; whereas before matching, 87.1% of patients on larotrectinib had a performance score of 0 to 1.

Several tumor types were observed in the SOC arm, including uterine (4%), biliary (4%), stomach (4%), endometrial (4%), CUP (4%), breast (4%), salivary gland (7%), non–small cell lung cancer (18%), soft tissue sarcoma (21%), and CRC (32%). In the larotrectinib arm, there were no cases of uterine, biliary, stomach, or endometrial cancer. When matching the number of lines of therapy since the patients’ diagnosis, 71.4% of patients had 0-2 lines of therapy and 28.6% had 3 or more lines of therapy since the time of their diagnosis. Most patients (64.3%) had stage III or IV disease at their initial diagnosis, with 17.9% of patients initially diagnosed with stage 0 to II disease. Most patients also had no or unknown brain metastases.

Investigators noted some study limitations, including that MAIC assumes that all effect modifiers are considered, and this analysis was limited to data reported from 2021. However, Bokemeyer concluded there is a clear OS advantage for using larotrectinib in 

Sustained Efficacy Supports Use in CNS Tumors

Larotrectinib (Vitrakvi) continued to produce rapid and durable responses, with a high disease control rate (DCR) and an acceptable toxicity profile in patients with 

 fusion–positive, primary central nervous system (CNS) tumors, according to long-term data from 2 clinical trials presented during the 2022 ASCO Annual Meeting.

Data from the long-term analysis of those treated on the phase 1/2 SCOUT study (NCT02637687) and the phase 2 NAVIGATE study (NCT02576431) showed that larotrectinib elicited an objective response rate (ORR) of 30% (95% CI, 16%-47%). In 28 pediatric patients, the ORR was 39% (95% CI, 22%-59%). The best responses included complete response (CR) in 8% of patients, partial response (PR) in 22%, stable disease (SD) for 24 weeks or more in 43%, SD for less than 24 weeks in 14%, and progressive disease (PD) in 14%.

A DCR for 24 weeks or more was observed in 73% of patients (95% CI, 56%-86%); in adult patients, this rate was 24% (95% CI, 14%-79%) and in pediatric patients, it was 82% (95% CI, 63%-94%). At a median follow-up of 25.6 months, the median DOR was not reached (95% CI, 4.9-not evaluable [NE]). The 24-month DOR was 53% (95% CI, 23%-83%).

Data from SCOUT and NAVIGATE served as partial basis for an accelerated FDA approval granted in 2018 for larotrectinib as treatment of adult and pediatric patients with solid tumors that have a 

 gene fusion without a known acquired resistance mutation.

A prior report from these trials showed that larotrectinib achieved an ORR of 30% (95% CI, 16%-49%) and a 24-week DCR of 73% (95% CI, 54%-87%) in 33 evaluable adult and pediatric patients with TRK fusion primary CNS tumors. In addition, treatment with larotrectinib resulted in tumor shrinkage in 82% of patients who had measurable disease.

The dataset for the long-term analysis included 26 pediatric patients from SCOUT who were under age 21 years with locally advanced metastatic solid tumors or CNS tumors and 12 adult and adolescent patients from NAVIGATE who were 12 years of age or older with advanced solid tumors and TRK fusion cancer. Collectively, the 38 patients were evaluated for the primary end point of investigator-assessed ORR per RECIST 1.1 criteria. The secondary end points assessed for the long-term analysis included duration of response (DOR), progression-free survival (PFS), overall survival (OS), and safety.

Adult patients were administered oral larotrectinib at 100 mg twice daily, and pediatric patients received 100 mg/m

 twice daily. Patients enrolled had a median age of 10.8 years (range, 1.3-79.0) and 50% were female. The most common type of 

 gene fusion identified in patients at baseline was 

 in 74%, followed by NTRK1 in 16%, and 

 in 11%. ECOG performance score ranged from 0 to 2; most patients (53%) had a score of 0, 34% had a score of 1, and 11% had a score of 2.

Most patients assessed had high-grade glioma (61%) and low-grade glioma (24%). Sixteen percent of patients had other histologies.

In terms of prior therapies, 71% of patients had undergone surgery, 58% received radiotherapy, and 87% had systemic therapy, and the median number of prior systemic therapies was 1 (range, 0-8). In fact, the majority of patient assessed had a median of 1 prior therapy (42%), whereas 21% had 2, another 21% had 3 or more, and 16% had 0.

On prior therapy, 6% of patient achieved a CR and 3% achieved a PR. The most common response to prior therapy was SD in 36%. Moreover, 30% of patients had PD during prior therapy, and 24% had other responses.

Data from SCOUT and NAVIGATE showed that treatment with larotrectinib lasted for 0.1 months to 38.7 months. The median time to response was 1.9 months (range, 1.0-3.8). Fifty-eight percent of patients progressed on larotrectinib and 4 continued the treatment post-progression for 4 weeks or more.

The median follow-up for PFS assessment was 27.4 months, and larotrectinib achieved a median PFS of 16.5 months (95% CI, 6.7-NE). At 24 months, the PFS rate observed with larotrectinib was 40% (95% CI, 24%-57%). At a median follow-up of 26.7 months, the median OS was not reached (95% CI, 22.6-NE). The 24-months OS rate among the patients evaluated was 65% (95% CI, 47%-83%).

Fifty-five percent of patients in the dataset experienced treatment-related adverse effects (TRAEs), but 86% of them were grade 1 or 2. Eight percent of patients experience grade 3 and 4 TRAEs, which included gamma-glutamytransferase increase, hyperglycemia, hypernatremia, hyponatremia, and neutrophil count decrease. The most common TRAEs of any grade included upper respiratory tract infection, vomiting, diarrhea, alanine aminotransferase increase, headache, anemia, and aspartate aminotransferase increase. 

Treatment-related neurologic AEs were also observed in the cohort. Notably, most events were grades 1 and 2 and occurred in the same patients. However, memory impairment was observed in 2 patients, 1 of whom was an adult and other a pediatric patient. No patients in the study discontinued treatment because of a TRAE.

As data support the use of NTRK inhibitors in practice, diagnostic approaches to identify these fusions should become part of routine practice. Testing for 

 fusions can be executed using a variety of methods including immunohistochemistry, fluorescence in situ hybridization, reverse transcriptase polymerase chain reaction and next-generation sequencing. Each of these methods provides advantages and disadvantages in terms of cost, array of detectable mutations and alterations, and availability.

Bokemeyer C, Paracha N, Lassen U, et al. Overall survival (OS) of patients with TRK fusion–positive cancer receiving larotrectinib versus standard of care (SoC): A matching-adjusted indirect comparison (MAIC) using real-world data (RWD). 

. 2022;40(suppl 16):6579. doi:10.1200/JCO.2022.40.16_suppl.6579

Vitrakvi. Prescribing information. Bayer HealthCare Pharmaceuticals Inc; 2021. Accessed June 24, 2022. bit.ly/3A7EfH7

Demetri G, Peters S, Davies J, et al. Characteristics and outcomes of patients (pts) with NTRK fusion-positive (NTRK+) metastatic / locally advanced (LA) solid tumours receiving non-TRK inhibitor (TRKi) standard of care (SoC), and prognostic value of NTRK fusions in clinical practice. 

. 2021;32(suppl_5):S382-S406. doi:10.1016/ annonc/annonc686

Perreault S, Drilon AE, Lassen LN, et al. Long-term control and safety of larotrectinib in a cohort of adult and pediatric patients with tropomyosin receptor kinase (TRK) fusion primary central nervous system (CNS) tumors. 

. 2022; 40 (suppl 16):2010. doi:10.1200/ JCO.2022.40.16_suppl.2010

Doz F, van Tilburg, Geoerger B, et al. Efficacy and safety of larotrectinib in TRK fusion-positive primary central nervous system tumors. 

. 2022;24(6):997-1007. doi:10.1093/neuonc/ noab274

Penault-Llorca F, Rudzinski ER, Sepulveda AR. Testing algorithm for identification of patients with TRK fusion cancer. 

. 2019;72(7):460-467. doi:10.1136/jclinpath-2018-205679

The role of NTRK gene fusions across multiple cancer types have led investigators to confirm the marker in real-world data analyses.

Real-World Data Underscore Importance of NTRK Gene Fusions in Tumor Agnostic Therapies 

Recent FDA approvals of several tumor-agnostic drugs are shifting the cancer treatment paradigm from traditional organ/ histology-directed therapies to a precision cancer medicine approach using biomarkers to guide treatment.

 One such biomarker is oncogenic fusions, the detection and targeting of which has been hailed as one of the most important breakthroughs in oncology over the past century.

Patients with non–small cell lung cancer (NSCLC) are among the major beneficiaries of this approach. Currently targetable oncogenic fusions are relatively rare in NSCLC; they are found in less than 1% to approximately 5% of tumors, depending on the fusion type and population. Nevertheless, the development of therapies targeting these oncogenic drivers has been associated with improved outcomes for molecularly selected subsets of patients with NSCLC, which remains the second most common cancer worldwide and the leading cause of cancer death.

, a panel of lung cancer experts discussed advances in targeting oncogenic fusions in NSCLC, including NTRK gene fusions, ALK rearrangements, and ROS1 rearrangements. Starting in 2011, the FDA has approved therapies targeting these fusions in first- and second-line settings for populations that include patients with NSCLC (FIGURE).

“We’re going to be entering a multidimensional universe of biomarkers for each patient. This isn’t going to be a binary world of EGFR mutant or wild-type, but it’s going to be secondary mutations, immunophenotypic markers that are complementing each other,” Martin F. Dietrich, MD, PhD, said.

The expansion of molecular markers shows the growing importance of next-generation sequencing, panelist Timothy Craig Allen, MD, JD, emphasized. “It is becoming impossible—if not already impossible—to find everything that’s out there just using [testing] kits or immunohistochemistry,” he said. “There is still, perhaps, a role for immunohistochemistry in [certain] settings, but I don’t think that excludes a full next-generation sequencing profile on our patients. We never know what we’re going to see and, as we know, even a small percentage of patients with lung cancer with a certain mutation, [considering] all the patients we have, there are going to be a reasonably large number of [patients] around the country who could benefit from [targeted] therapy.”

 gene fusions encode TRK fusion proteins, which are oncogenic drivers in various tumor types,” Misako Nagasaka MD, PhD, explained. TRKs are expressed in normal human neuronal tissue, where they are activated by neurotrophins to facilitate the development and function of the human nervous system. Three transmembrane proteins encoded by 3 

 gene fusions with a variety of partner genes have been observed in human cancers, which can result in an altered TRK fusion protein with enhanced kinase function that promotes cancer cell growth and survival. In NSCLC, 

 gene fusions have been observed in less than 1% of tumors.

Two FDA-approved tyrosine kinase inhibitors (TKIs) have shown significant benefit in targeting TRK fusion–positive cancers, including NSCLC: larotrectinib (Vitrakvi) and entrectinib (Rozlytrek).

 Entrectinib is also approved for ROS1-positive NSCLC.

During the 2021 American Society of Clinical Oncology (ASCO) Annual Meeting, investigators presented an updated efficacy and safety analysis of the 3 pivotal larotrectinib trials, LOXO-TRK-14001 (NCT02122913), NAVIGATE (NCT02576431), and SCOUT (NCT02637687) that led to larotectinib’s approval.

 The analysis included data for 206 patients (median age, 38 years) with 21 different solid tumors.

With a median follow-up of 22.3 months, the objective response rate (ORR) remained 75% (95% CI, 68%-81%), with a complete response (CR) in 22%, partial response (PR) in 53%, stable disease (SD) in 16%, and progressive disease (PD) in 6%. In 15 patients with brain metastases, the ORR was 73% (95% CI, 45%-92%), most of which were PRs (73%). There were no CRs. Among all patients, the median duration of response (DOR) was 49.3 months (95% CI, 27.3-not estimable [NE]). The median progression-free survival (PFS) was 35.4 months (95% CI 23.4-55.7), and the median overall survival (OS) was not reached; however, 77% of patients were alive at 36 months.

“Treatment-related adverse events [TRAEs] were mainly grade 1 and 2, with 18% having grade 3 or 4 TRAEs. Only 2% of patients discontinued treatment due to TRAEs,” Nagasaka said.

The benefit from larotrectinib therapy was maintained in extended follow-up data from an expanded population of 244 participants in the same pivotal studies with 25 different tumor types, including 9% with lung cancer, according to findings presented at the 2022 ASCO Annual Meeting. Among the expanded cohort, the ORR was 69% (95% CI, 63%-75%), with CRs in 21%, PRs in 43%, SD in 17%, and PD in 8%. Among 18 patients with baseline central nervous system (CNS) metastases, the ORR was 83% (95% CI, 59%-96%). After a median follow-up of at least 28.3 months, the median DOR for all evaluable patients was 32.9 months (95% CI, 27.3-41.7), the median PFS was 29.4 months (95% CI, 19.3-34.3), and median OS was not reached.

Nagasaka noted that efficacy and safety findings for patients with NTRK fusion–positive lung cancer were comparable to data for the overall population, according to an analysis of participants in the NAVIGATE and LOXO-TRK-14001 trials.

 At 2022 ASCO, investigators presented updated findings for 26 patients with lung cancer included in those studies, including 24 who had adenocarcinoma histology, one with atypical carcinoid, and one with neuroendocrine histology.

 gene fusions were observed in 21 patients (81%) and 

Among 23 evaluable patients, larotrectinib therapy demonstrated an ORR of 83% (95% CI, 61%-95%), with 2 CRs, 17 PRs, 4 SD, and no PD. Of 10 evaluable patients with CNS metastases, the ORR was 80% (95% CI, 44%-97%), with no CRs, 8 PRs, 2 SD, and no PD. After follow-up of at least 12.9 months, the median for all evaluable patients was not reached for DOR (95% CI, 9.5-NE) or PFS (95% CI, 9.9-NE). The median OS was 40.7 months (95% CI, 19.4-NE).

Most TRAEs were grade 1 or 2. Five patients reported grade 3 TRAEs of increased alanine and aspartate aminotransferase, myalgia, hypersensitivity, and weight gain. No patients discontinued treatment because of TRAEs.

Investigators presented updated efficacy and safety data during the ASCO 2020 Virtual Scientific Program for patients treated with entrectinib during 3 phase 1/2 studies, ALKA372-001 (EudraCT 2012-000148-88), STARTRK-1 (NCT02097810), and STARTRK-2 (NCT02568267), whose findings paved the way for its FDA approval. The analysis included 74 evaluable patients with advanced or metastatic NTRK fusion–positive solid tumors, 13 of which were NSCLC.

The ORR was 63.5% (95% CI, 51.5%74.4%) in all evaluable patients, 63.8% (95% CI, 50.1%-76.0%) in patients without baseline CNS disease, and 62.5% (95% CI, 35.4%-84.8%) in patients with baseline CNS disease. Among all evaluable patients, the median DOR was 12.9 months (95% CI, 9.3-NE), the median PFS was 11.2 months (95% CI, 8.0-15.7), and the median OS was 23.9 months (95% CI, 16.0-NE).

“Safety was in line with what was previously reported. The most common grade 3 or higher TRAEs were weight gain, anemia, and fatigue,” Nagasaka said. Weight gain and anemia both occurred in 7.1% of patients and fatigue in 6.2% of patients.

The latest data for a larger cohort of patients with 

 fusion–positive solid tumors treated during the same clinical studies supported those findings, according to results presented at 2022 ASCO.

 Among 150 evaluable patients, the ORR was 61.3% (95% CI, 53.1%-69.2%), including CRs in 16.7%, PRs in 44.7%, SD in 10.0%, PD in 12.0%, missing or unevaluable scans for 12.0%. The median DOR was 20.0 months (95% CI, 13.2-31.1), the median PFS was 13.8 months (10.1-20.0), and the median OS was 37.1 months (95% CI, 27.2-NE).

Responses were observed for a range of tumor types. In 31 patients with NSCLC, the ORR was 64.5% (95% CI, 3.7-58.8). Among 31 patients with baseline CNS metastases across tumor types, the ORR also was 61.3% (95% CI, 42.2%-78.2%), including 2 patients with CRs, 17 with PRs, 4 with SD, and 2 with PD. The median DOR in the CNS cohort was 17.2 months (95% CI, 9.0-33.3), median PFS was 11.7 months (95% CI, 4.9-30.3), and median OS was 20.0 months (95% CI, 7.9-NE). 

No new safety signals were observed in the safety population of 235 patients. TRAEs resulted in discontinuation in 7.2% of participants.

, the authors stated that TRK inhibitors improve responses and survival for eligible patients vs historical outcomes with non–TRK-targeted standard therapies across most tumor histologies including in later lines of therapy where TRK inhibitors are likely to be used.

In NSCLC, larotrectinib and entrectinib elicited a more robust response in patients with 

 gene fusions than would be expected with standard therapies,

 a finding that persisted in updated data presented at 2022 ASCO (TABLE

 However, Carlson et al noted that “matching-adjusted indirect comparison and simulation studies suggest greater benefit to survival outcomes with larotrectinib compared with entrectinib.”

Dietrich agreed with this assessment. “They’re both good drugs, but larotrectinib seems to be doing better in the lung space in both progression and OS. For familiarity, and because of those data pieces, that might be my preferred choice,” he said. Fernando C. Santini, MD, also agreed. “TKI therapies should be the first-line approach for all patients with 

 fusion–positive cancer. Larotrectinib is probably a little more active and powerful. The response rates are amazing, along with the duration of response activity in the brain. It’s a no-brainer,” he said.

In patients experiencing progression on larotrectinib, Dietrich recommends rebiopsying and resequencing. “Many of these tumors remain dependent on the NTRK fusion,” he said, adding that several novel TRK inhibitors may offer promise in this setting, such as LOXO-195, repotrectinib, and taletrectinib. Of these, he said LOXO-195 is no longer in development but is still available on compassionate use through the manufacturer, whereas repotrectinib and taletrectinib are currently being explored in clinical trials.

Repotrectinib (TPX-0005) is a small macrocyclic TKI of ROS1, TRK, and ALK designed to overcome resistance mutations.

 In May 2022, the FDA granted repotrectinib a breakthrough therapy designation as a treatment for patients with 

-positive metastatic NSCLC who have previously been treated with a ROS1 TKI but have not received prior platinum-based chemotherapy. The drug previously gained breakthrough therapy status for 2 additional indications: patients with 

-positive metastatic NSCLC who have not been treated with a ROS1 TKI and patients with advanced solid tumors with an 

 gene fusion who have progressed after 1 or 2 prior TRK TKIs, with or without prior chemotherapy, and have no satisfactory treatment options.

Investigators are studying repotrectinib in the phase 1/2 TRIDENT-1 trial (NCT03093116) in patients with advanced solid tumors harboring 

 Assessment of the early phase 2 TRIDENT-1 dataset that included the first 39 treated patients showed an ORR of 50% (3/6 patients; 95% CI, 12%-88%) among the 

 The ORR was 67% (4/6 patients; 95% CI, 22%-96%) among 

-positive patients previously treated with 1 prior TKI but no prior platinum-based chemotherapy, whereas it was 40% (2/5 patients each; 95% CI, 5%-85%) among both 

A panel of experts discuss their perspectives on testing and targeted therapy for actionable mutations in non–small cell lung cancer.

Vol. 23/No. 13 | Oncology Live®

Machine Learning Develops Into a Powerful Tool in Cancer Care

Expanding Opportunities for Real-World Evidence in Oncology

Real-World Data Underscore Importance of NTRK Gene Fusions in Tumor Agnostic Therapies

A Growing Arsenal of NSCLC Treatments Take Aim at Oncogenic Fusions

More Novel Therapies Are on the Horizon in Multiple Myeloma